Conference
Deterministic Feature Decoupling by Surfing Invariance Manifolds
| العنوان: | Deterministic Feature Decoupling by Surfing Invariance Manifolds |
|---|---|
| المؤلفون: | Martínez-Enríquez, Eduardo, Portilla, Javier |
| المساهمون: | Ministerio de Economía y Competitividad (España) |
| بيانات النشر: | Institute of Electrical and Electronics Engineers |
| سنة النشر: | 2020 |
| المجموعة: | Digital.CSIC (Consejo Superior de Investigaciones Científicas / Spanish National Research Council) |
| مصطلحات موضوعية: | Nonlinear orthogonal features, Sample statistics, Regression, Manifolds, Decoupled features, Orthokurtosis |
| الوصف: | ICASSP 2020 - 2020 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP), Barcelona, Spain, 4-8 May 2020 . -- 5 pags., 1 fig. ; We introduce a formalism that justifies and extends a heuristic method for algebraically decoupling deterministic features that recently proved useful for improving feature-based classification. Our new formalism is based on defining transformations inside manifolds, by following trajectories along the features' gradients. Through these transformations we define a normalization that, we demonstrate, allows for decoupling differentiable features. By applying this to sampling moments, we obtain a quasi-analytic solution for the orthokurtosis, a modification of the kurtosis that is not just decoupled from mean and variance, but also from skewness. After theoretically motivating feature decoupling for random data distributions, we illustrate with a regression problem example how decoupled features may perform significantly better than coupled ones. ; Funded by the Spanish Government grant FIS2016-75891-P |
| نوع الوثيقة: | conference object |
| اللغة: | unknown |
| تدمد: | 0736-7791 |
| Relation: | #PLACEHOLDER_PARENT_METADATA_VALUE#; info:eu-repo/grantAgreement/MINECO/Plan Estatal de Investigación Científica y Técnica y de Innovación 2013-2016/FIS2016-75891-P; Postprint; http://dx.doi.org/10.1109/ICASSP40776.2020.9053597; Sí; Proceedings - ICASSP, IEEE International Conference on Acoustics, Speech and Signal Processing 2020: 6049-6053 (2020); http://hdl.handle.net/10261/230640; http://dx.doi.org/10.13039/501100003329 |
| DOI: | 10.1109/ICASSP40776.2020.9053597 |
| DOI: | 10.13039/501100003329 |
| الاتاحة: | http://hdl.handle.net/10261/230640 https://doi.org/10.1109/ICASSP40776.2020.9053597 https://doi.org/10.13039/501100003329 |
| Rights: | open |
| رقم الانضمام: | edsbas.B9F1DEB3 |
| قاعدة البيانات: | BASE |
| تدمد: | 07367791 |
|---|---|
| DOI: | 10.1109/ICASSP40776.2020.9053597 |